cd /news/ai-ethics/96-on-the-midterm-48-on-the-final-ho… · home topics ai-ethics article
[ARTICLE · art-45299] src=gadgetreview.com ↗ pub= topic=ai-ethics verified=true sentiment=↓ negative

96% on the Midterm, 48% on the Final: How a Professor Trapped 50 AI Cheaters Using Simple Math

Brown University economics professor Roberto Serrano caught over 50 students cheating on a take-home midterm exam after the class average hit 96%, compared to 48% on the in-person final. Serrano offered the take-home format to ease trauma from a campus shooting, but students used AI to answer questions. The university has remained largely silent, and its academic code still does not explicitly address AI use.

read3 min views1 publishedJun 30, 2026
96% on the Midterm, 48% on the Final: How a Professor Trapped 50 AI Cheaters Using Simple Math
Image: Gadgetreview (auto-discovered)

A professor who never offered a take-home exam in 34 years of teaching finally gives one — out of compassion for students traumatized by a campus shooting. The class average comes back: 96 out of 100. Roughly 40 students score a perfect 100. On an advanced mathematical economics exam where he deliberately altered the theoretical assumptions from class. If that sounds implausible, the data gets worse. This isn’t just a cheating scandal. It’s a stress test for whether elite universities will confront AI-enabled fraud or politely pretend it didn’t happen.

The Numbers Don’t Lie – Even When Students Do #

A game theory specialist inadvertently designed the perfect natural experiment to catch AI cheaters.

Professor Roberto Serrano, Harrison S. Kravis University Professor of Economics at Brown and a game theory specialist,

offered ECON 1170’sMarch 5 midterm as a take-home, closed-book exam. The course usually draws under 30 students. This semester, 86 enrolled — likely attracted by the format. Teaching assistants flagged answer passages matching ChatGPT outputs almost verbatim, according to reporting from Gigazine and Marginal Revolution.

Serrano didn’t void the midterm. Instead, he warned students the in-person final would count heavily if grade distributions looked suspicious. The results told a clear story:

  • The final’s class average landed at 48 out of 100 - Of 27 students who skipped the final entirely, 22 had scored perfect 100s on the midterm
  • Over 50 students were flagged with what Serrano calls “overwhelming” evidence of AI use

The midterm-to-final collapse isn’t a statistical anomaly — it’s a pattern with a very obvious explanation.

The emotional context sharpens the betrayal. Serrano offered the take-home format after a December 2025 campus shooting killed two students, including Ella Cook, who had just met him to discuss becoming his advisee in economics and mathematics. It was his first take-home exam in three and a half decades of teaching, offered specifically to ease students’ anxiety about being on campus.

“Academic integrity is a value worth defending. The faculty cannot be left on its own in a battle that is decisive if we want to preserve the future of higher education,” Serrano stated publicly.

The Sound of Administrative Silence #

Brown’s leadership went quiet when confronted with evidence — and the university’s academic code still doesn’t mention AI.

Serrano escalated to Brown’s president and dean. The response, he says, was “cold” silence. Only after he formally filed with the Standing Committee for the Academic Code did administrators issue a note calling the incident a “wake-up call” — a phrase that lands somewhere between understatement and institutional deflection, given the scale of alleged fraud.

Brown’s academic code still doesn’t explicitly define permitted AI use, according to Associate Dean Love Wallace. The committee evaluates evidence of a student’s writing process rather than relying on automated detection — a reasonable approach, given that widely used

toolslike GPTZero hover around 60% accuracy. That’s barely better than a coin flip when the stakes involve academic expulsion.

The broader picture isn’t encouraging. Princeton recently ended its 133-year tradition of unproctored, honor-code exams, partly in response to AI cheating concerns. Journalist Theo Baker captured the scale of the problem in The New York Times: “AI has made deception easier and more remunerative than ever before,” adding that he doesn’t know “a single person who hasn’t used AI to get through some assignment in college.”

Serrano — who lost his sight at 17 and relies on assistive technology himself — knows precisely where the line falls between tools that support learning and tools that replace it. He has now vowed to:

  • Eliminate take-home exams permanently
  • Stop counting weekly problem sets toward final grades

The real question is whether Brown will match that clarity before the next scandal makes the choice unavoidable.

── more in #ai-ethics 4 stories · sorted by recency
── more on @brown university 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/96-on-the-midterm-48…] indexed:0 read:3min 2026-06-30 ·